The Business Challenge
The business situation the organisation faced before the project began.
The workshop took bookings by phone and walk-in only. Customers searching online could not find them for the faults they actually had, and leads that came in after hours were lost.
Why AI Was the Right Solution
Why AI was the right approach — and what alternatives were considered.
AI was used to build, not to run the workshop. Directing AI tools let one person scope, build, test and maintain a full booking system on a small-business budget. The running system is deliberately plain: no AI makes decisions about customers or bookings, so the owner can trust exactly what it does.
How the Solution Was Delivered
Discovery, design, development, testing and rollout — the journey, not the tooling.
- 1Sat with the owner and listed how a booking really happens today
- 2Built the booking flow and the staff dashboard
- 3Wrote security tests before release, each proven to fail when its protection is removed
- 4Fixed what the tests found, including a real flaw in how returning customers were stored
- 5Deployed to Johannesburg and checked every page live
- 6Set up Search Console and weekly monitoring
Key Technical & Architecture Decisions
Architecture, model selection, workflow and trade-offs.
Hosted in Johannesburg so customer data stays in South Africa, on an encrypted volume with daily backups.
A new booking links to an existing customer record and never overwrites it.
The production server refuses to start if any secret is still a placeholder.
Security headers on every page, checked live on three hostnames.
WhatsApp for alerts because the owner already lives in it; no new app to learn.
Challenges & How They Were Solved
Obstacles hit along the way and how they were overcome.
A real flaw found before launch: an anonymous booking could overwrite a returning customer's details. Fixed so a new booking links to the record and never edits it, and a test now guards it.
Tests passing on the wrong setup: the laptop ran older library versions than the live server, so a passing test proved less than it seemed. Fixed by testing on a clean copy built from the production requirements.
User Adoption & Change Management
How users responded — training, change management and feedback loops.
The owner needs nothing new to know a booking arrived: it comes to his WhatsApp. Getting staff into the habit of moving each booking along in the dashboard is still in progress, and is the main adoption work left.
Business Outcomes & Impact
The measurable outcomes from the underlying AI Deployment — with the story behind them.
591 Google search impressions and 25 clicks in 28 days (average position 5.1, 87 queries), from a workshop that had no online booking before.
The story behind the numbers
The owner now gets a WhatsApp message the moment a booking arrives, instead of finding out when he is back at the desk.
Google Search Console, 28 days to 5 October 2026: 591 impressions, 28 clicks, 4.7% click rate, average position 5.1, across 90 search queries.
What drives those results so far, measured: people searching the workshop's name, and 'near me' searches for steering and suspension repair. The fault-specific service pages have not yet appeared in search results; that is the next thing to measure.
The site gets found because a real business with real customers stands behind it, and the site tells Google clearly what and where it is.
Lessons Learned
What surprised the team, what worked well, and what would be done differently.
- Test on exactly what production runs
- A protection is not proven until its test has failed once
- A real local business behind the site is what gets it found first
Future Opportunities
Where this solution could go next.
Track month by month which service pages start appearing in search.
A simple monthly report to the owner: bookings, searches, calls.
Staff habits in the dashboard, so every booking has a status.
Professional Reflections
Guidance for another builder tackling a similar problem.
Test on exactly what production runs, and write the test that proves a protection works before you trust it. Then measure where search traffic actually lands, not where you hoped it would. Here the honest answer was the business name and 'near me' searches, not the content pages, and that is worth knowing early.